the next generation of data products | anacondacon 2017
TRANSCRIPT
#OpenDataScienceMeans #AnacondaCON
What is a data product?
Goodbye, big data!
Hello, AI!
Data products are products that would not be possible without the
use of data.
(it’s all the same.)
What makes a great data product?
(it’s boring)
(thisisnotasolvedproblem)
What’s different in 2017?
0 to 1 matters0 to 1000000 matters
1) layers of abstraction
[anaconda]
2) natural language interfaces
3) data!
+ high value problems
(don’t forget!)
First, automate things we pay people to do today.
Then, build new products that were simply never viable before.
[caution]
Unfortunately, you can’t just buygood data products.
There is a generic formulation of your problem.
Then there’s your problem.
This is the data product gap.
It’s hard, and sometimes it doesn’t work.
There are unpredictable edge cases.
Organizations are not designed for data science or for developing data products.
Use an ExperimentalDevelopment Process.
Find the simplest possible algorithm that will work at scale.
Have a plan for operationalization and maintenance.
There’s a new practice for the UX of ML.
Academic ResearchStartups
Enterprise
Here’s where we look…
1) a researchbreakthrough
2) a change in economics
http://www.mkomo.com/cost-per-gigabyte
3) a capability becomes a commodity
4) new data is available
Examples
The real impact will be in making complex data simple.
There’s been an increase in sales!
Moving beyond counting words to computable representations of
concepts.
Tools for combining domain knowledge with small data to make
risky decisions.
Hilary Mason@hmason
#OpenDataScienceMeans #AnacondaCON